extraction, analysis, atom mapping, classification
and naming of reactions from pharmaceutical elns
Roger Sayle1, Daniel Lowe1, Noel O’Boyle1, Mick Kappler2,
3
4
5
Anna Paola Pelliccioli , Nick Tomkinson and Daniel Stoffler
NextMove Software Ltd, Cambridge, UK. 2 Hoffmann-La Roche, Nutley, USA. 3 Novartis NIBR, Basel, Switzerland.
4 AstraZeneca R&D, Alderley Park, UK. 5 F. Hoffmann-La Roche, Basel, Switzerland.
Abstract
1. Overview
Electronic Laboratory Notebooks (ELNs) are widely used in the pharmaceutical
industry for recording the details of chemical synthesis experiments. The primary
use of this information is often for the capture of intellectual property for future
patent filings, however this data can also be used in a number of additional
applications, including synthetic accessibility calculations, reaction planning and
reaction yield/optimization. Not only does a pharmaceutical ELN capture those
classes of reactions suitable for small scale medicinal chemistry, but it is also
uniquely a source of information on failed and poor yield reactions; an important
source of data rarely found in the scientific literature or commercial reaction
databases.
2. Method overview (workflow)
In this work we describe the use of a suite of programs for extracting reaction
information and associated textual and numeric data from the PKI/CambridgeSoft
eNotebook ELN, converting and normalizing it to “open” flat files that can be read
by third party applications, such as Accelrys/MDL RD files or reaction SMILES.
Accelrys
Pipeline Pilot
(AstraZeneca, AbbVie
& Hoffmann-La Roche)
HazELNut
Perkin Elmer Informatics
(formerly CambridgeSoft)
eNotebook v9, v11 or v13
Oracle Server
version 10 or 11
Filbert
NameRXN
Microsoft Windows or Linux
Cobnut
6. NameRXN: Reaction naming and classification
One useful form of reaction analysis is to recognize each experiment in the ELN as
an instance of a named or known reaction, such as a Diels –Alder cycloaddition,
Suzuki coupling or chiral separation. The NameRXN program uses a dictionary of
SMIRKS-like transformations to annotate reactions in RD, SD, RXN or reaction
SMILES format with a name, a reaction category and where applicable an
identifier into the Royal Society of Chemistry’s RXNO ontology.
The top five reactions in a major pharmaceutical company’s ELN were:
#1
#2
#3
#4
#5
ID
1.3.1
7.1
1.7.6
2.1.2
2.2.3
Reaction Name
Buchwald-Hartwig amination
Nitro to amino
Williamson ether synthesis
Carboxylic acid + amine reaction
Sulfonamide Shotten-Baumann
Reaction Category
N-arylation with Ar-X
Nitro to amine reduction
O-substitution
N-acylation to amide
N-sulfonylation
RXNO #
0000192
0000337
0000090
0000165
Using the reaction classification categories of Carey et al.[1] the contents of the
ELN may be presented as a pie-chart of the kinds of transformations it contains.
ChemAxon
JChem Cartridge
(GlaxoSmithKline
& Novartis)
Elsevier Reaxys
(Hoffmann-La Roche)
3. HazELNut:
HazELNut: Reaction
Reaction export
export from
from PKI/CambridgeSoft
PKI/CambridgeSoft eNotebook
eNotebook ELN
ELN
3.
The primary step in exporting a reaction database from an ELN is performed by
HazELNut, which interprets the chemist’s hand-drawn sketch into a connection
table. For RD and SD formats, this includes writing the textual, numeric, tabular
and molecular data as tagged fields in the output file. The process can work
either from XML files exported from the client, or more typically by querying the
Oracle server via OCI. A large pharmaceutical ELN can be exported overnight (or
transatlantic over a weekend), though incremental export allows a day’s or week’s
experiments to be written in a few minutes. Experimental write-ups (and other
rich text) are converted from RTF to text or HTML, superatoms are expanded,
labels preserved, and so on. The images below show an ELN reaction, and some
of the exported fields as they appear in the corresponding RD file format output.
$DTYPE HEADER:EXPERIMENT.HEADER:CREATION.DATE
$DATUM 22-Oct-2010
$DTYPE HEADER:EXPERIMENT.HEADER:CREATION.TIME
$DATUM 19:58:18 -0500
$DTYPE DISCOVERY.CHEMISTRY:REACTANTS(1):CHEMICAL.STRUCTURE
$DATUM c1ccc(cc1)C(=O)O
$DTYPE DISCOVERY.CHEMISTRY:REACTANTS(1):NAME
$DATUM benzoic acid
$DTYPE DISCOVERY.CHEMISTRY:REACTANTS(1):MOLECULAR.WEIGHT:VALUE
$DATUM 122.12
$DTYPE DISCOVERY.CHEMISTRY:PRODUCTS(1):NAME
$DATUM N-benzyl-N-ethylbenzamide
$DTYPE DISCOVERY.CHEMISTRY:PRODUCTS(1):CHEMICAL.STRUCTURE
$DATUM CCN(Cc1ccccc1)C(=O)c2ccccc2
$DTYPE DISCOVERY.CHEMISTRY:PRODUCTS(1):MOLECULAR.FORMULA
$DATUM C16H17NO
$DTYPE DISCOVERY.CHEMISTRY:PRODUCTS(1):MOLECULAR.WEIGHT:VALUE
$DATUM 239.31
$DTYPE DISCOVERY.CHEMISTRY:PRODUCTS(1):YIELD:VALUE
$DATUM 89.8%
$DTYPE DISCOVERY.CHEMISTRY:PREPARATION
$DATUM In a 5 mL round-bottomed flask, benzoic acid (200 mg, 1.64 mmol
N-benzylethanamine (266 mg, 1.97 mmol, Eq: 1.2) and HATU (747 mg, 1.97
mmol, Eq: 1.2) were combined with DMF (5 ml) to give a light brown
solution. Hunig's Base (423 mg, 572 μl, 3.28 mmol, Eq: 2.0) was added.
The reaction mixture was heated to 50 oC (Pressure: 1012 mbar, Rxn
Molarity: 328 mM, Reaction Time: 12 min, Molarity Entered?: false) and
stirred. The crude material was purified by flash chromatography
7. Third-party atom
Atom mapping
Mapping
One way to investigate the quality of the reaction sketches in an ELN is to apply an
automatic atom-atom mapping algorithm, and assess the fraction of unmapped
product atoms and average number of carbon-carbon bonds broken. At the Fall
2012 ACS meeting in Philadelphia, we presented an initial comparison of thirdparty atom mapping software for the purpose of identifying incorrectly drawn
100
reactions.
Here we summarize recent
improvements to this
approach, using consensus
methods to combine the
results of multiple atom
mapping algorithms, and
combining atom-mapping
with reaction naming for
ELN quality assurance.
90
Percent of reactions with all product atoms mapped
1
Marvin 5.12
80
ChemDraw 12
Indigo 1.1
70
Indigo 1.1 (lenient)
60
ICMap 5.10
PipelinePilot
50
MDL Cheshire
40
Verified/Recognised by
NameRXN
(62%)
30
20
10
0
Atom mapping algorithms alone
4. Filbert:
Filbert: Reaction
Reaction file
file format
format conversion
conversion
4.
Whilst HazELNut can export reactions in a variety of file formats, its goal is to
“dump” the raw data; processing and file format conversion is performed by
Filbert. This program interconverts MDL RXN, RD and SD file formats, ISIS Sketch,
CDXML and reaction SMILES. The roles of agents, catalysts and solvents drawn
above and below a reaction arrow can be preserved using ChemAxon’s RXN file
extensions, stripped or treated as reactants. Co-ordinates can optionally be
centered, rescaled or regenerated algorithmically. Atom maps can be removed.
Molecules from the reactant/agent tables can be added to the sketch if missing.
5. Cobnut: MDL/Accelrys
MDL/Accelrys RDF
RDF file
file format
format tag
tag stripping/renaming
stripping/renaming
Typically, ELNs are heavily customized with custom experiment types to capture
the data needs of their target scientists. The raw export of this data results in RD
files where the field/tag names for yields, volumes, chemist’s names,
experimental write-ups, etc. vary from reaction to reaction. The utility Cobnut
was implemented to simplify and speed-up the process of normalizing (renaming)
RD tags from different data sources, and stripping out those that aren’t required.
Consensus Result
Combined with NameRXN
8. Summary
and future work
Conclusions
The use of the HazELNut suite of tools allows the synthetic chemistry data locked
up in corporate ELNs to be exploited in scientifically novel ways. Current efforts
include optimization of reaction conditions by plotting yields against catalyst and
solvent in SpotFire, and research to reduce the number of steps and application of
low yield reaction strategies by identifying/reusing in-house insights/expertise.
9. Bibliography
1. John S. Carey, David Laffan, Colin Thomson and Mike T. Williams, “Analysis of
the Reactions used for the Preparation of Drug Candidate Molecules”,
Organic & Biomolecular Chemistry, Vol. 4, pp. 2337-2347, 2006.
2. Stephen D. Roughley and Allan M. Jordan, “The Medicinal Chemist's Toolbox:
An Analysis of Reactions Used in the Pursuit of Drug Candidates”, Journal of
Medicinal Chemistry, Vol. 54, 3451-3479, 2011.
3. Mikko J. Vainio, Thierry Kogej and Florian Raubacher, “Automated Recycling of
Chemistry for Virtual Screening and Library Design”, Journal of Chemical
Information and Modeling (JCIM), Vol. 52, No. 7, pp. 1777-1786, June 2012.
www.nextmovesoftware.co.uk
www.nextmovesoftware.com
NextMove Software Limited
Innovation Centre (Unit 23)
Cambridge Science Park
Milton Road, Cambridge
England, UK CB4 0EY